Re: [PATCH] ioctl_tty.2: Add example how to get or set baudrate on the serial port

["Alejandro Colomar \(man-pages\) via Libc-alpha" <libc-alpha@sourceware.org>]

Hi Pali,


On 7/30/21 2:05 PM, Pali Rohár wrote:

> 
>>> +
>>> +int
>>> +main(int argc, char *argv[])
>>> +{
>>> +#if !defined(TCGETS2) || !defined(TCSETS2) || !defined(BOTHER)
>>> +    fprintf(stderr, "TCGETS2, TCSETS2 or BOTHER is unsupported\\n");
>>> +    return 1;
>>> +#else
>>
>> Do we want the program to compile if those are unsupported?
> 
> My intention was to provide example which compiles fine on any Linux
> architecture and in case of error it reports it at runtime.
> 
> On architectures where are TCGETS2/TCSETS2 ioctls unsupported, there are
> still supported TCGETS/TCSETS ioctls. So proper Linux portable program
> should fallback to TCGETS/TCSETS ioctls with Bnnn constants.
> 
> So for example setting baudrate to 115200 would be possible via
> predefined constant B115200 in c_cflag member even when struct termios2
> with c_ospeed is unsupported.
> 
> I just did not put this fallback into this example as it would be quite
> loooong (as it is needed to add big switch for every Bnnn constant and
> convert numeric value into Bnnn) and example show how to use
> TCGETS2/TCSETS2 (not TCGETS/TCSETS).

Okay, I leave it up to you what you consider best to do :)

> 
>> Maybe you can #error there and simplify the reader having to parse the
>> preprocessor directive mentally:
>>
>> #if !defined...
>> # error ...
>> #endif
>>
>> I know it's non-standard, but I think it's common enough so that we can use
>> it here.
> 
> #error is standard. It is already defined in C99 (section 6.10.5 Error
> directive).

Ahh, it is #warning that is non-standard!  Thanks.  I forgot that.

> 
>>> +    struct termios2 tio2;
>>> +    int fd, rc;
>>> +
>>> +    if (argc != 2 && argc != 3) {
>>> +        fprintf(stderr, "Usage: %s device [new_baudrate]\\n", argv[0]);
>>
>> We use \e for printing the escape character.  Not \\
> 
> Ok!
> 
>> CC: Branden
>>
>> See groff_man(7):
>>     Portability
>>         [...]
>>
>>         Similar  caveats  apply  to escapes.  Some escape sequences
>>         are however required for correct typesetting  even  in  man
>>         pages and usually do not cause portability problems:
>>
>>         [...]
>>
>>         \e     Widely used in man pages to  represent  a  backslash
>>                output  glyph.  It works reliably as long as the .ec
>>                request is not used, which should  never  happen  in
>>                man pages, and it is slightly more portable than the
>>                more exact ‘\(rs’  (“reverse  solidus”)  escape  se‐
>>                quence.
>>
>>
>>> +        return 1;
>>> +    }
>>> +
>>> +    fd = open(argv[1], O_RDWR | O_NONBLOCK | O_NOCTTY);
>>> +    if (fd < 0) {
>>> +        perror("open");
>>> +        return 1;
>>
>> exit(EXIT_FAILURE);
>>
>>> +    }
>>> +
>>> +    rc = ioctl(fd, TCGETS2, &tio2);
>>> +    if (rc) {
>>> +        perror("TCGETS2");
>>> +        close(fd);
>>> +        return 1;
>>
>> exit(3)
>>
>>> +    }
>>> +
>>> +    printf("%u\\n", tio2.c_ospeed);
>>
>> \e
>>
>>> +
>>> +    if (argc == 3) {
>>> +        tio2.c_cflag &= ~CBAUD;
>>> +        tio2.c_cflag |= BOTHER;
>>> +        tio2.c_ospeed = tio2.c_ispeed = atoi(argv[2]);
>>> +
>>> +        rc = ioctl(fd, TCSETS2, &tio2);
>>> +        if (rc) {
>>> +            perror("TCSETS2");
>>> +            close(fd);
>>> +            return 1;
>>
>> exit(3)
>>
>>> +        }
>>> +    }
>>> +
>>> +    close(fd);
>>> +    return 0;
>>
>> exit(3)
> 
> Interesting... Do you prefer to use exit(EXIT_SUCCESS) instead of return 0?

I don't care in my own code.
I typically use return 0 at the end of main.
But the historical convention in manual pages is using exit(EXIT_SUCCESS),
so let's follow that :)


Tanks,

Alex


-- 
Alejandro Colomar
Linux man-pages comaintainer; https://www.kernel.org/doc/man-pages/
http://www.alejandro-colomar.es/